Acoustical cues versus top-down bias in infants' parsing

Abstract

French liaison involves the surfacing of an underlying consonant as the onset of the following vowel-initial word (e.g., les amis - /le/ /zami /), creating misalignment. However, acoustic cues that support vowel-initial parsing may exist. In a preferential looking procedure we examined French-learning 30-month-olds' parsing in liaison-related cases. Familiarization sentences in Experiment 1 contained a determiner preceding a vowel-initial non-word (e.g., ces onches). Two test conditions followed. The vowel-initial condition presented the vowel-initial non-word versus another non-target (onches - èque). The consonant-initial condition tested the consonant-initial parse (zonches - zèque). Infants in the vowel-initial, but not the consonant-initial condition, showed discrimination (p=.008), i.e., they correctly parsed the vowel-initial target, possibly using acoustic cues. Knowledge of underlying liaison consonants can also explain these results. In Experiment 2 we removed acoustic cues to vowel-initial parsing by using a consonant-initial non-word following a determiner in familiarization sentences (e.g., un zonche). Test conditions were the same as those in Experiment 1. Infants yielded the same results as Experiment 1, showing discrimination only in the vowel-initial condition (p=.047). Taken together, 30-month-olds perceived /z/ as an independent element unrelated to the preceding word; they used this partial liaison knowledge, rather than possible acoustical cues, for parsing. © 2013 Acoustical Society of America.